To start on a positive note, HubSpot explicitly states that customer data is NOT used for AI training within ChatGPT when using this connector. In fact, both HubSpot and OpenAI have put policies in place to safeguard this.
OpenAI's Stance: For paid plans OpenAI does not use your business data (inputs or outputs) to train their models by default. Your data is encrypted both at rest and in transit.
HubSpot's Stance: While HubSpot can use your data to develop their own Breeze AI features (with an opt-out available), for this specific connector with ChatGPT, the commitment is no data usage for OpenAI's model training.
That being said, we’ve seen instances before where so-called “protected” data was later revealed by an AI chatbot. So, what should you still be mindful of?
1. Data Usage vs. Model Training:
Even though data isn't used for training, it is being processed by ChatGPT to answer our queries. This means data is temporarily in OpenAI's environment. While they have strong security protocols, it's good to understand the exact processing and retention specifics—which are usually covered in their enterprise agreements—before giving it access to sensitive customer data via the connector.
2. HubSpot Permissions:
The connector only has access to the data the user who set it up is permitted to see within HubSpot, which is a critical layer of control...unless every user in your portal is a Super-Admin. If someone has broad access in HubSpot, they will have broad access via the connector. So, robust internal access management is super important.
While no system is foolproof, the explicit commitment from both HubSpot and OpenAI not to use our CRM data for general model training should at least give you some air cover for sensitive data conversations.
Personally, I think the data privacy ship sailed long ago and, whether it is stored in HubSpot’s servers or OpenAIs, there is always a significant risk of a data leak. It’s just a part of operating in a technology-driven world.
You should exercise sound judgement and take proactive steps to improve your data security, but if you completely avoid the use of AI for fear of what will happen to your data, you are going to fall behind your competition. AI can be a powerful tool, and with careful usage and good internal data governance, go-to-market teams can use it to gain a ton of value from it.
---
I just shared 6 go-to-market use cases for HubSpot's Deep Research connector—including sample prompts—covering marketing, sales, and RevOps in my latest Hubsessed newsletter. Check it out at the link under my name at the top of this post.
This post was originally shared by on Linkedin.